A multistep mechanism for the activation of ... - Semantic Scholar

4 downloads 0 Views 450KB Size Report
histones using the drug Trichostatin A. This same combination of demethylation and histone acetylation is also required to activate germline transcription and ...
A multistep mechanism for the activation of rearrangement in the immune system Yanhong Ji*, Jianmin Zhang*, Alfred Ian Lee†, Howard Cedar*, and Yehudit Bergman*‡ *Departments of Experimental Medicine and Cellular Biochemistry, Hebrew University Medical School, Jerusalem 91120, Israel; and †Section of Immunobiology, Yale University School of Medicine, New Haven, CT 08360 Communicated by Gary Felsenfeld, National Institutes of Health, Bethesda, MD, May 1, 2003 (received for review August 7, 2002)

L

ymphocyte development is characterized by the assembly of B and T cell antigen receptors. Variable region gene segments are flanked by conserved recombination signal sequences (RSSs) that serve as a substrate for two lymphoid specific proteins, Rag1 and Rag2 (1–3), which together with other enzymes mediate rearrangement. Despite the fact that all these rearrangement reactions are mediated by universal RSSs on the DNA and are carried out by the same set of enzymes, recombination takes place in a programmed, cell type-specific, locussequential manner (3, 4), suggesting that this regulation is carried out at the level of locus accessibility. Although little is known about the mechanisms that repress unscheduled gene rearrangement, one of the main players in this process must be DNA methylation (5). In the case of the mouse ␬ locus, there is clear-cut evidence that demethylation actually precedes and may be required for normal recombination to occur in vivo during B cell development (6). In keeping with this concept, it has been demonstrated that DNA methylation can inhibit rearrangement both in cell culture (7, 8) and in transgenic mice (9), possibly through its effect on histone acetylation and chromatin structure (10, 11). We used a simple recombination substrate in a stable transfection system to demonstrate that initially the presence of DNA methylation serves to establish a repressed chromatin structure characterized by DNaseI insensitivity and local histone H4 deacetylation, as well as H3 methylation on lysine 9. Removal of this methylation does not itself release the lock on recombination, but it does make the locus more susceptible to histone modification, which can activate transcription and recombination. The endogenous ␬ locus is also subject to a similar control mechanism. These results serve as a model for understanding both the general repression of rearrangement in nonlymphoid cells as well as the multilayered sequential process of chromatin opening that occurs during lymphoid differentiation. Materials and Methods Plasmid. pMX-RSS-EGFP was constructed by inserting the 0.8-kb fragment containing the enhanced GFP (EGFP) isolated from pEGFP (CLONTECH) flanked by the 12- and 23-RSSs into pMX (12). The EGFP is in the reverse orientation relative to the 5⬘ LTR. The murine stem cell virus (MSCV)-Rag2-IRESwww.pnas.org兾cgi兾doi兾10.1073兾pnas.0932635100

Rag1 vector was kindly provided by Sara Cherry (Harvard Medical School, Cambridge, MA). The Rag2 ribosomal entry site (IRES)-Rag1 cassette was subcloned into MSCV (13) and cut with EcoRI兾SacI, a digest that removes the phosphoglycerate kinase promoter driving the neo gene. Cell Culture and Drug Treatment. The 38B9 mouse pre-B cells (14)

were treated with 10 ng兾ml Trichostatin A (TSA) (Sigma) for 24 h before examining the degree of V(D)J recombination or local transcription. p53⫺/⫺ and p53⫺/⫺兾Dnmt1⫺/⫺ fibroblasts (described in Supporting Materials and Methods, which is published as supporting information on the PNAS web site, www. pnas.org) were treated with 50 ng兾ml TSA for 24 h to study ␬ germ-line transcription and Rag-induced double-strand breaks (DSBs). Methylation in Vitro and Transfection. 38B9 cells (1 ⫻ 107) were

stably electroporated with methylated or unmethylated linearized plasmid DNA as described (ref. 15 and Supporting Materials and Methods). p53⫺/⫺ and p53⫺/⫺兾Dnmt⫺/⫺ cells were retrovirally infected with the pMSCV-Rag2-IRES-Rag1 retroviral vector for 5 h as described (16), and 24 h later TSA (50 ng兾ml) was added for an additional 24 h. Recombination Assay. Genomic DNA (1 ␮g) was subjected to PCR amplification using primers P1 and P3 for the first round, and P2 and P4 for the second round (see Fig. 2 and Table 1, which is published as supporting information on the PNAS web site), which are able to amplify molecules that underwent V(D)J recombination only. DNaseI Sensitivity Assay. Cells (2 ⫻ 108) from methylated-EGFP-

containing 38B9 clones were mixed with an equal number of cells from either unmethylated or demethylated clones. In some experiments, three to seven individual clones from each category were mixed together to generate pools. DNaseI sensitivity assay was performed as described (17).

Chromatin Immunoprecipitation (ChIP) Assay. ChIP analysis was performed as described (18) by using rabbit polyclonal antibodies directed against acetylated histone H4 (10 ␮g per 60 ␮g of mononucleosomal DNA), or antibodies against Me-H3(K9) (20 ␮l per 10 ␮g of mononucleosomal DNA) (Upstate Biotechnology, Lake Placid, NY), followed by protein A-Sepharose (5 mg per 60 ␮g of mononucleosomal DNA) (Sigma). EGFP molecules were amplified with P6 and P7, ␣-actin with ␣-actinR and ␣-actinL, and ␤-actin with ␤-actinR and ␤-actinL (Supporting Materials and Methods and Table 1). RT and Ligation-Mediated PCR (LM-PCR). Total RNA was collected

from the cells (TRIzol, Sigma) and reverse transcribed (Pro-

Abbreviations: RSS, recombination signal sequence; EGFP, enhanced GFP; TSA, Trichostatin A; LM-PCR, ligation-mediated PCR; ChIP, chromatin immunoprecipitation; DSB, doublestrand breaks. ‡To

whom correspondence should be addressed. E-mail: [email protected].

PNAS 兩 June 24, 2003 兩 vol. 100 兩 no. 13 兩 7557–7562

BIOCHEMISTRY

Rearrangement of immune receptor loci is a developmentally controlled process that takes place exclusively in lymphoid cells. We have used a stable transfection system in pre-B cells to show that DNA methylation brings about histone underacetylation, histone H3(K9) methylation, DNaseI resistance, and strong inhibition of both transcription and recombination. Strikingly, this repression is maintained in dividing cells even after removal of the original methyl groups responsible for its establishment, but in this state, rearrangement can now be induced by reacetylation of local histones using the drug Trichostatin A. This same combination of demethylation and histone acetylation is also required to activate germline transcription and recombination from the endogenous ␬ locus in vivo. These results indicate that the regulation of rearrangement is carried out by a multilayered synergistic process.

Fig. 1. Methylation status of the integrated EGFP reporter gene. (A) A map of the EGFP recombination reporter gene. EGFP (speckled line) is in inverse orientation with regard to the 5⬘ LTR (dark line), flanked by 12 RSS (open triangle) and 23 RSS (filled triangle). Restriction enzyme sites for EcoRI (R), HpaII (Hp), HhaI (Hh), NcoI (N), and SacI (S) are marked. The line below the restriction map represents the EcoRI–NcoI fragment used as a radioactive probe. (B) Methylation pattern of the transfected EGFP reporter gene. The EGFP reporter gene was methylated in vitro by using HhaI and HpaII methylases, and both the modified and unmodified constructs were cotransfected with a plasmid carrying the neor selection gene into 38B9 cells. Neomycinresistant colonies were expanded and then used to prepare DNA for Southern blot analysis. DNA from unmethylated (um), methylated (m), and demethylated (dm) clones (cl) was digested with NcoI兾EcoRI with or without HhaI and HpaII and was subjected to blot hybridization using the EGFP radioactive probe (A). Similar results were obtained from all of the unmethylated (26), methylated (25), and demethylated (7) clones used in this study. Although this analysis examines only the methylation state of a subset of sites within the EGFP regions, all of the HpaII and HhaI sites shown in the figure were assayed by methyl-sensitive PCR and shown to be ⬍10% methylated in the demethylated clones.

mega), and RT-PCR analysis of unrearranged EGFP molecules was performed by using P1 with P8 and P2 with P7 (see Fig. 2 Upper). RT-PCR analysis of ␬ germ-line transcription was carried out by using primers GL␬1 and GL␬2 for detecting the 0.8kb RNA, and GL␬3 and GL␬4 for the 1.1-kb RNA. ␤-actin (using ␤-actin1 and ␤-actin2, Table 1) and adenine phosphoribosyltransferase (Aprt) (using Aprt1 and Aprt2) gene products were used to measure the amount of cDNA in the RT-PCR reactions. Both RT-PCR conditions were the same as described above for the ␤-actin gene in the ChIP assay. Rag1 and -2 RNA molecules were measured by using Rag1L and Rag1R, Rag2L and Rag1R (Table 1). LM-PCR was done as described (ref. 19 and Supporting Materials and Methods), by using BW1 and BW2 primers (Table 1). Genomic DNA (2 ␮g from RAG-infected p53⫺/⫺ and p53⫺/⫺兾Dnmt1⫺/⫺ fibroblasts) was subjected to linker ligation followed by PCR using primers L␬1 and BW3 for detection of J␬1 signal ends (see Fig. 6 and Table 1). PCR products were analyzed by Southern blots using the radioactive L␬5 oligonucleotide as a probe. Results Methylation Inhibits EGFP Reporter Gene Rearrangement. To study

V(D)J recombination, we used a reporter plasmid composed of an EGFP gene sequence flanked by two RSSs. This DNA was methylated in vitro by using HhaI and HpaII methylases, and both the modified and unmodified constructs were stably cotrans-

7558 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0932635100

Fig. 2. DNA methylation represses EGFP reporter gene rearrangement. A schematic diagram of EGFP reporter gene rearrangement is shown. The arrows represent the locations of primers. Genomic DNA from um-cl1, dm-cl2, and m-cl3 was amplified by a first set of primers (1;3) and then by a nested set of primers (2;4) to detect rearrangement. The ␣-actin gene was used as a control for the amount of input DNA. Similar results were obtained for all of the methylated, unmethylated, and demethylated clones used in this study.

fected individually into the 38B9 pre-B cell line, which expresses Rag1 and Rag2 activities. Southern blot analysis of DNA from representative colonies (see Supporting Materials and Methods) revealed that in most cases the methylated construct remained modified, the unmethylated EGFP stayed unmodified (Fig. 1B), and these patterns were stable for many generations. We found, however, that several of the methylated-EGFP-containing clones underwent extensive spontaneous demethylation after several months of growth in culture (Fig. 1B; see Table 2, which is published as supporting information on the PNAS web site). To measure the potential of each type of DNA to undergo rearrangement in vivo, special oligonucleotides were designed for detecting rearranged molecules by PCR analysis (Fig. 2). Unmethylated transfected genes undergo a high degree of active recombination. In contrast, no rearranged products were obtained with DNA from clones transfected with methylated EGFP constructs (Fig. 2 and Table 2), and dilution analysis indicated that we could have detected rearrangement at levels 10,000-fold less than that observed for the unmethylated DNA (data not shown). Because only ⬇10% of the CpG residues (HpaII and HhaI sites) on the test vector were actually methylated in these experiments, it is possible that this effect is even stronger in vivo, where the density of methylation may be higher. These results are consistent with previous experiments in other systems that also show that DNA methylation inhibits recombination (7–9). In striking contrast to this strong correlation, no rearrangement was observed in clones that had undergone spontaneous demethylation (Fig. 2 and Table 2). To demonstrate that the integrated demethylated DNA itself had not lost its intrinsic ability to undergo rearrangement, we amplified the EGFP reporter gene from one of the demethylated lines by PCR, cloned the products into a new plasmid vector, and then retransfected this DNA in unmethylated form back into wild-type 38B9 cells. Similar transfections were also carried out with total genomic DNA isolated from dm-clone2, by using Neo selection to identify those cell colonies that acquired the EGFP vector region. After both of these rescue procedures, the EGFP reporter DNA underwent rearrangement at about the same level as the original unmethylated vectors (Fig. 7, which is published as supporting information on the PNAS web site). This repreJi et al.

have been removed. Similar results were obtained by using pools of methylated or demethylated clones (data not shown and Table 2). These results may explain the inaccessibility of these templates to Rag1 and Rag2 and thus their inability to undergo rearrangement. Histone Modification and V(D)J Recombination. We next tested

sents clear-cut proof that the demethylated templates had not undergone any change in primary sequence that could prevent rearrangement, and it demonstrates functionally that the repressive state can be maintained even though the DNA has lost the ability to reestablish repression. In light of these results, we conclude that the inability of the spontaneously demethylated EGFP molecules to undergo rearrangement is a result of chromatin structural features that were initially set up by methyl groups on the transfected DNA and apparently remain in place, even after demethylation. DNaseI Sensitivity of the Transfected EGFP Recombination Reporter Genes. Because previous studies demonstrated that DNA meth-

ylation plays a role in the establishment of inactive chromatin (20–23), we next analyzed DNaseI sensitivity for each substrate. To be able to directly compare active and inactive templates in a single experiment, cells from methylated clones were mixed with cells from either unmethylated or demethylated clones before the preparation of nuclei and DNaseI treatment. The digested DNA was extracted and cut with EcoRI兾SacI in the presence or absence of HhaI to distinguish between methylated and unmethylated EGFP molecules and was then subjected to blot hybridization (Fig. 3A). As shown in Fig. 3B, unmethylated exogenous DNA is more sensitive to DNaseI than its methylated counterpart. However, DNA that was initially methylated and in a closed chromatin conformation remains inaccessible to DNaseI even after the methyl groups themselves

Ji et al.

Activation of Rearrangement by TSA. The fact that methylated DNA is packaged into a nucleosome structure containing underacetylated histones suggests that this feature participates in the prevention of gene rearrangement. To test whether this is the case (8, 29), we treated cells with the histone deacetylase inhibitor, TSA (31), and then examined the degree of recombination. When TSA was added to cells containing the EGFP constructs, neither unmethylated nor methylated DNA were affected. Unmethylated DNA retained its normal degree of rearrangement, whereas methylated DNA remained resistant to recombination (Fig. 5A) even after the addition of high concentrations of TSA (data not shown). Strikingly, however, DNA substrates that had undergone spontaneous demethylation responded to TSA with at least a 5,000-fold increase in rearrangement, reaching a level similar to that observed for the original unmethylated DNA clones (Fig. 5A and Table 2). These results PNAS 兩 June 24, 2003 兩 vol. 100 兩 no. 13 兩 7559

BIOCHEMISTRY

Fig. 3. DNaseI sensitivity of the methylated, unmethylated, and demethylated EGFP reporter genes. (A) Cells from methylated EGFP-containing clones were mixed with cells from unmethylated or demethylated clones before nuclei preparation and subsequent DNaseI treatment. The digested DNA was further cut with EcoRI兾SacI and HhaI, to distinguish between methylated and unmethylated EGFP molecules, and subjected to blot hybridization using the radioactive probe indicated in Fig. 1 A. (B) The graph (based on phosphorimaging of the blot) shows EGFP reporter gene DNaseI sensitivity of methylated, unmethylated, and demethylated clones. Similar results were obtained by using pools of methylated or demethylated clones (data not shown and Table 2).

whether histone modification (24–28) may be responsible for the methylation effects seen in the above experiments. To this end, we examined the histone H4 acetylation state of nucleosomes in the EGFP region by using a ChIP assay. Mononucleosomes were first isolated from stably transfected cell lines and anti-acetylated (Ac) H4 antibodies were used to specifically precipitate the highly acetylated fraction. DNA was isolated from the unbound and bound mononucleosomes taken from individual transfected clones and tested for sequence content by the PCR (Fig. 4A). For each cell line, we first showed that ␤-actin gene sequences are enriched as compared with a nonexpressed gene (␣-actin) (29). When the transfected sequences were assayed in these same mononucleosomal DNA fractions, we found that the unmethylated EGFP copies were highly enriched (20-fold) in the bound fraction (Fig. 4C) and thus must be packaged in an open chromatin structure. In contrast, methylated EGFP sequences were enriched for acetylation to a much lesser degree. Similar results were also obtained when this experiment was done on pooled, as opposed to individual, colonies (Fig. 4C and Table 2). We next carried out ChIP on mononucleosomes isolated from transfected 38B9 cells by using an antibody specific for MeH3(K9), which is associated with inactive genes (30). In this case, the inactive ␣-actin gene is highly enriched in the bound fraction as compared with ␤-actin sequences. Strikingly, we found that the unmethylated EGFP transgene is packaged with nucleosomes carrying relatively unmethylated H3(K9) (less than ␣actin, for example), whereas methylated DNA is enriched for Me-H3(K9) in this assay (Fig. 4 B and C). It should be noted that these effects of DNA methylation on histone modification are probably not just an indirect consequence of local transcription, because even nontranscribed bacterial sequences behave in a similar manner when introduced into cells in culture (data not shown). Thus, taken together, our results show that the presence of DNA methylation directly affects local histone modification states, which may in turn be involved in modulating accessibility to the rearrangement machinery. In light of these results, one might have expected that EGFP sequences would undergo repackaging to an active nucleosome structure after spontaneous demethylation. ChIP analysis, however, clearly shows that the core histones present on this DNA actually retain their inactive state, with H4 being relatively unacetylated and H3 still highly methylated at the lysine 9 residue (Fig. 4C). Thus, the ability to undergo rearrangement appears to be directly correlated with chromatin structure.

Fig. 4. Histone modification of EGFP reporter genes. Micrococcal nuclease-prepared mononucleosomes from individual clones um-cl1, dm-cl2, and m-cl3 or from pools (pl) of unmethylated (um), demethylated (dm), and methylated (m) clones were precipitated with antibodies directed against acetylated histone H4 (A) or methylated histone H3(K9) (B). DNA was extracted from unbound and bound fractions and subjected to quantitative PCR using primers (6;7) (see Fig. 2) for the EGFP region and for mouse ␤- and ␣-actin as controls. Three concentrations (1-, 3-, and 9-fold) are shown. ChIP of Ac-H4 was also carried out on pooled colonies (data not shown). (C) The data in A and B have been redrawn in graphic form. For experiments on Ac-H4, the degree of enrichment (bound兾unbound) is normalized to that obtained for the ␣-actin gene in each cell type, whereas experiments on Me-H3(K9) were normalized to that obtained for the ␤-actin gene. It should be noted that identical results were obtained for the LTR region of the vector.

indicate that the inactive state can be maintained for many generations in culture even in the absence of the original methyl moieties responsible for its establishment. Activation of Transcription by TSA. Because the EGFP construct

contains an intact LTR viral promoter, we were able to determine whether the reporter genes used in our experiment also undergo RNA synthesis in conjunction with the ability to undergo rearrangement. When we used a RT-PCR assay (Fig. 5B), we detected a strong signal from unmethylated reporter sequences tested in several different individual clones (Table 2). In contrast, transcription from methylated DNA was strongly repressed (⬇5,000-fold). We also analyzed a number of clones that underwent spontaneous demethylation to varying degrees. Strikingly, these templates remained almost totally inactive (Fig. 5B). With the addition of TSA, however, transcription was increased to control levels (Fig. 5B), even though this drug had almost no effect on methylated templates (Table 2). When treated cells were grown in the absence of TSA for many generations, all resulting subclones retained a high activity level (Fig. 5C), suggesting that histone deacetylation itself may be required for maintenance of the inactive state. Furthermore, these findings also confirm that the original repression seen for this transgene must have been caused by DNA methylation and was not a result of intrinsic silencing at the site of integration. To test whether the endogenous ␬ locus behaves in a similar manner, we used a p53⫺/⫺兾Dnmt1⫺/⫺ fibroblast cell line developed in our laboratory (Supporting Materials and Methods). Unlike in wild-type or p53⫺/⫺ fibroblasts, almost all (95%) of the CpG residues in these cells are unmethylated (data not shown), even though these cells were actually descended from embryos that must have undergone methylation at the time of implantation (see Supporting Materials and Methods). As shown in Fig. 6A, no ␬ germ line (GL␬) transcription could be detected in either 7560 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0932635100

cell type, even by using sensitive RT-PCR analysis. In contrast, RNA synthesis from both the proximal (0.8-kb transcript) and the distal upstream promoter (1.1-kb transcript) (32) was strongly induced by treatment with TSA, but only from the unmethylated cells (Fig. 6A). This may represent a more general phenomenon, because other methylated genes apparently behave in a similar manner (33). In light of these results, we then asked whether the endogenous ␬ locus can also be made available for rearrangement in nonlymphoid cells. To this end, we introduced expression vectors for the Rag1 and Rag2 genes into p53⫺/⫺ and p53⫺/⫺兾Dnmt1⫺/⫺ cells, and we verified by using RT-PCR that these genes are indeed expressed after transfection (Fig. 6B). Rearrangement potential was determined by assaying for DSB at the J␬1 RSS using LM-PCR. Strikingly, specific DSBs were observed exclusively from the unmethylated templates, and only after TSA treatment. Thus, in a manner similar to the reporter transgenes, the endogenous ␬ locus appears to be repressed synergistically by both DNA methylation and histone deacetylation. When taken together, these experiments reaffirm our initial observation that the inactive state can be maintained in growing cells even in the absence of methylation. Discussion Our results suggest that DNA methylation mediates rearrangement repression through several independent mechanisms. On the one hand, the presence of methylation clearly brings about local changes in histone modification (22, 23), and this may be, in part, what causes methylated DNA templates to be packaged into a DNaseI-inaccessible conformation (29). On the other hand, there is no question that histone deacetylation alone cannot account for all of the effects of DNA methylation, because TSA-mediated reacetylation was unable to induce transcription and rearrangement on a methylated template, even Ji et al.

BIOCHEMISTRY

Fig. 5. TSA treatment activates rearrangement and transcription of the demethylated EGFP reporter gene. (A) Genomic DNA was prepared from um-cl1, dm-cl2, and m-cl3 cells that were treated with or without TSA (10 ng兾ml for 24 h), and then amplified with nested primers (1;3 and 2;4, see Fig. 2) that detect rearranged (Re) molecules. ␣-Actin was used as a control for measuring the amount of input DNA. Similar results were obtained for other unmethylated (4), demethylated (6), and methylated (11) clones (Table 2). (B) RT-PCR analysis of EGFP transcription in unmethylated, demethylated, and methylated clones with (⫹) or without (⫺) TSA treatment. Amplification was done with nested primers (1;8 and 2;7, see Fig. 2). The ␤-actin gene was used as a control for these experiments. Similar results were obtained for other unmethylated (4), demethylated (6), and methylated (5) clones (Table 2). (C) dm-cl2 was single-cell cloned to obtain dm-cl2⬘. Treated dm-cl2⬘ cells were grown in the absence of TSA, subcloned, and then examined for EGFP transcription by RT-PCR as above.

though this same exact treatment was shown to be effective after spontaneous removal of the original methyl groups (Figs. 5 and 6). It is interesting to note that although TSA works primarily by inhibiting histone deacetylase, this may be sufficient to bring about more comprehensive changes in chromatin structure by inducing DNaseI sensitivity (29) and by chemically preventing methylation at histone H3 lysine 9. Because most of the RSSs do not contain any CpG residues, it is unlikely that methylation acts directly to prevent the binding of rearrangement factors. Our results not only provide information for understanding how rearrangement is inhibited in nonlymphoid cells, but also shed light on the mechanisms that may be used to release the antigen receptor loci from repression to allow programmed recombination at its proper time during lymphoid development. The J␬ region, for example, is both methylated and packaged within an inactive histone structure before rearrangement (6, 27), and each of these impediments must presumably be removed during B cell development. Our experiments on transgenes further suggest that this process may be carried out by at least two independent mechanisms, and it is likely that this is also the case in vivo during normal development, because repression of the endogenous ␬ locus is also carried out by a synergistic mechanism (Fig. 6). It has already been demonstrated that immune-specific enhancers are required for demethylation of the receptor loci (15, 34–36), and it is likely that enhancers also play a role in local Ig histone deacetylation (10, 24, 37, 38). Indeed, these cis-acting sequences have been shown to be essential for both acetylation and rearrangement independent of their ability to bring about demethylation (10, 24, 39), and even forced undermethylation by itself is not sufficient to induce recombination (16). Conversely, several studies have suggested that treatment with drugs that inhibit histone deacetylases may be able to overcome the inhibitory effects of DNA methylation, but in these cases, the authors did not rigorously rule out the possibility that recombination actually occurred on a small number of unmethylated molecules (8, 25). It should be noted that although our studies show that both demethylation and histone acetylation are necessary for rearrangement, it is not clear how these events take place in vivo. It is certainly possible that demethylation occurs first. However, it is equally likely that enhancer-mediated histone acetylation is

Fig. 6. TSA treatment activates the endogenous ␬ gene. (A) Map of the ␬ locus showing the promoters (arrows) of the 1.1-kb (5⬘) and 0.8-kb (3⬘) germ-line transcripts, the positions of J␬1–5, and the constant region (C␬). Shown is RT-PCR analysis of p53⫺/⫺ and p53⫺/⫺兾Dnmt1⫺/⫺ fibroblasts for the 0.8- and 1.1-kb Ig␬ germ-line (GL␬) transcripts with (⫹) or without (⫺) TSA treatment. The mouse Aprt gene is used as a control for these experiments. (B) p53⫺/⫺ or p53⫺/⫺兾Dnmt1⫺/⫺ cells with or without transfected Rag expression vectors were treated with TSA and assayed by RT-PCR for Rag1 and Rag2 expression using ␤-actin as a control. LM-PCR (see diagram) was used to detect DSB at the J␬1 RSS using ␣-actin as a control for DNA content. Southern blot analysis using four different methyl-sensitive restriction enzymes showed that the ␬ gene region is unmethylated in Dnmt1⫺ cells (data not shown).

Ji et al.

PNAS 兩 June 24, 2003 兩 vol. 100 兩 no. 13 兩 7561

moieties can actually be maintained on its own, even after the methyl groups have been removed. This finding suggests that there must be a methylation-independent mechanism for genocopying epigenetic states onto newly made DNA molecules during replication. Although this phenomenon has been observed in other organisms (43, 44), this represents the first demonstration of chromatin maintenance in mammals.

the initial trigger and that this is followed by removal of the methyl groups (40), which then leads to recombination. One of the most fascinating findings in this article is that transfected DNA is able to maintain an inactive chromatin structure even in the absence of the original signals involved in the establishment of the repressed state. In the presence of DNA methylation, local histones alter their modification state through reactions that are probably mediated by methyl binding protein complexes that recruit histone modification enzymes such as deacetylases (23, 41) and methylases (42). Because cells already have a simple mechanism for maintaining DNA methylation through replication, it is easy to understand how the inactive chromatin state is perpetuated from one cell generation to the next. Surprisingly, our results show that the inaccessible chromatin conformation initially laid down by the presence of methyl

We thank Dr. Sara Cherry for the MSCV-Rag2-IRES-Rag1 vector and the members of our laboratories for stimulating discussions. This work was supported by grants from the Israel Academy of Sciences (to Y.B. and H.C.), the German Israel Foundation (to Y.B.), the National Institutes of Health (to Y.B. and H.C.), the European Community Fifth Framework Quality of Life Program (to Y.B.), and the Israel Cancer Research Fund (to H.C.).

1. Schatz, D. G., Oettinger, M. A. & Baltimore, D. (1989) Cell 59, 1035–1048. 2. Oettinger, M. A., Schatz, D. G., Gorka, C. & Baltimore, D. (1990) Science 248, 1517–1523. 3. Gellert, M. (1996) Genes Cells 1, 269–275. 4. Lewis, S. A. (1994) Adv. Immunol. 56, 27–150. 5. Bergman, Y. & Mostoslavsky, R. (1998) Biol. Chem. 379, 401–407. 6. Mostoslavsky, R., Singh, N., Kirillov, A., Pelanda, R., Cedar, H., Chess, A. & Bergman, Y. (1998) Genes Dev. 12, 1801–1811. 7. Hsieh, C. L. & Lieber, M. R. (1992) EMBO J. 11, 315–325. 8. Cherry, S. R. & Baltimore, D. (1999) Proc. Natl. Acad. Sci. USA 96, 10788– 10793. 9. Engler, P., Haasch, D., Pinkert, C. A., Doglio, L., Glymour, M., Brinster, R. & Storb, U. (1991) Cell 65, 1–20. 10. Mathieu, N., Hempel, W. M., Spicuglia, S., Verthuy, C. & Ferrier, P. (2000) J. Exp. Med. 192, 625–636. 11. Whitehurst, C. E., Schlissel, M. S. & Chen, J. (2000) Immunity 13, 703–714. 12. Onishi, M., Kinoshita, S., Morikawa, Y., Shibuya, A., Phillips, J., Lanier, L. L., Gorman, D. M., Nolan, G. P., Miyajima, A. & Kitamura, T. (1996) Exp. Hematol. (Charlottesville, VA) 24, 324–329. 13. Grez, M., Akgun, E., Hilberg, F. & Ostertag, W. (1990) Proc. Natl. Acad. Sci. USA 87, 9202–9206. 14. Alt, F., Rosenberg, N., Lewis, S., Thomas, E. & Baltimore, D. (1981) Cell 27, 381–390. 15. Kirillov, A., Kistler, B., Mostoslavsky, R., Cedar, H., Wirth, T. & Bergman, Y. (1996) Nat. Genet. 13, 435–441. 16. Cherry, S. R., Biniszkiewicz, D., van Parijs, L., Baltimore, D. & Jaenisch, R. (2000) Mol. Cell. Biol. 20, 7419–7426. 17. Liu, J. K., Bergman, Y. & Zaret, K. S. (1988) Genes Dev. 2, 528–541. 18. Hebbes, T. R., Clayton, A. L., Thorne, A. W. & Crane-Robinson, C. (1994) EMBO J. 13, 1823–1830. 19. Schlissel, M., Constantinescu, A., Morrow, T., Baxter, M. & Peng, A. (1993) Genes Dev. 7, 2520–2532. 20. Keshet, E. & Cedar, H. (1983) Nucleic Acids Res. 11, 3571–3580. 21. Kass, S. U., Landsberger, N. & Wolffe, A. P. (1997) Curr. Biol. 7, 157–165. 22. Jones, P. L., Veenstra, G. J. C., Wade, P. A., Vermaak, D., Kass, S. U., Landsberg, N., Strouboulis, J. & Wolffe, A. P. (1998) Nat. Genet. 19, 187–191.

23. Nan, X., Ng, H.-H., Johnson, C. A., Laherty, C. D., Turner, B. M., Eisenman, R. N. & Bird, A. (1998) Nature 393, 386–389. 24. McMurry, M. T. & Krangel, M. S. (2000) Science 287, 49549–49558. 25. McBlane, F. & Boyes, J. (2000) Curr. Biol. 10, 483–486. 26. Kwon, J., Morshead, K. B., Guyon, J. R., Kingston, R. E. & Oettinger, M. A. (2000) Mol. Cell 6, 1037–1048. 27. Mae¨s, J., O’Neill, L. P., Cavelier, P., Turner, B. M., Rougeon, F. & Goodhardt, M. (2001) J. Immunol. 167, 866–874. 28. Agata, Y., Katakai, T., Ye, S. K., Sugai, M., Gonda, H., Honjo, T., Ikuta, K. & Shimizu, A. (2001) J. Exp. Med. 193, 873–880. 29. Eden, S., Hashimshony, T., Keshet, I., Cedar, H. & Thorne, A. W. (1998) Nature 394, 842–843. 30. Litt, M. D., Simpson, M., Gaszner, M., Allis, C. D. & Felsenfeld, G. (2001) Science 293, 2453–2455. 31. Yoshida, M., Horinouchi, S. & Beppu, T. (1995) BioEssays 17, 423–430. 32. Martin, D. J. & van Ness, B. G. (1990) Mol. Cell. Biol. 10, 1950–1958. 33. Cameron, E. E., Bachman, K. E., Myohanen, S., Herman, J. G. & Baylin, S. B. (1999) Nat. Genet. 21, 103–107. 34. Lichtenstein, M., Keini, G., Cedar, H. & Bergman, Y. (1994) Cell 76, 913–923. 35. Bergman, Y. (1999) Semin. Immunol. 11, 319–328. 36. Cedar, H. & Bergman, Y. (1999) Curr. Opin. Immunol. 11, 64–80. 37. Fernandez, L. A., Winkler, M. & Grosschedl, R. (2001) Mol. Cell. Biol. 21, 196–208. 38. Krangel, M. S. (2001) J. Exp. Med. 193, F27–F30. 39. Engler, P. & Storb, U. (2000) Mamm. Genome 11, 694–695. 40. Cervoni, N. & Szyf, M. (2001) J. Biol. Chem. 276, 40778–40787. 41. Jones, P. A. & Gonzalgo, M. L. (1997) Proc. Natl. Acad. Sci. USA 94, 2103–2105. 42. Fuks, F., Hurd, P. J., Wolf, D., Nan, X., Bird, A. P. & Kouzarides, T. (2002) J. Biol. Chem. 278, 4035–4040. 43. Cavalli, G. & Paro, R. (1999) Science 286, 955–958. 44. Nakayama, J., Klar, A. J. & Grewal, S. I. (2000) Cell 101, 307–317. 45. Howell, C. Y., Bestor, T. H., Ding, F., Latham, K. E., Mertineit, C., Trasler, J. M. & Chaillet, J. R. (2001) Cell 104, 829–838. 46. Carlson, L. L., Page, A. W. & Bestor, T. H. (1992) Genes Dev. 6, 2536–2541.

7562 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0932635100

Ji et al.